World indicators analysis
Ćukasz Kosturski, Krystian Moras
27 04 2022
library(tidyverse)
library(ggplot2)
library(BBmisc)
library("readxl")
# ggplot extensions
library(gganimate)
library(ggbump)
library(ggrepel)
if(!require(ggflags)) devtools::install_github("rensa/ggflags")
library(ggflags)
library(plotly)
library(countrycode) # library to extract codes, continents from country names
library(maps) # coordinates for the world map
library(RColorBrewer)
library(paletteer)
library("eurostat") #European borders and country codes
library("sf") # Simple features for plotting polygons
library(grid)
library(patchwork)
library(scatterpie) # distributed pie charts on the mapwdi <- read_excel('../data/World_Development_Indicators.xlsx')
oldnames <- colnames(wdi)
newnames <- gsub(pattern = '..YR[0-9][0-9][0-9][0-9].', '', colnames(wdi))indicators_base <- wdi %>%
rename_at(vars(oldnames), ~newnames) %>%
rename(series = "Series Name",country_name = "Country Name", country_code = "Country Code") %>%
filter(!country_name %in% c("Upper middle income","High income","Low & middle income","Low income","Middle income", "Data from database: World Development Indicators")) %>%
na_if("..")GDP per capita analysis
Top countries
This bump chart presents the ranking of countries with the highest GDP in years 2000 - 2020. We chose 10 countries with the highest mean rank within the given time interval.
GDP <- indicators_base %>%
filter( series == "GDP per capita (current US$)")%>%
filter(if_all('2000':'2020', ~ !is.na(.))) %>%
select(country_name, '2000':'2020') %>%
filter(country_name != "World") %>%
# single row - single observation - using tidyr
pivot_longer(cols = 2:22, names_to= 'year', values_to = 'GDP_per_capita') %>%
# transform datatypes
transform(GDP_per_capita = as.numeric(GDP_per_capita)) %>%
transform(year = as.numeric(year)) %>%
# ranking
group_by(year) %>%
mutate(rank = rank(-GDP_per_capita, ties.method = "random")) %>%
ungroup() %>%
# mean rank
group_by(country_name) %>%
mutate(mean_rank = mean(rank)) %>%
ungroup()
# take the best 10 countries over all years
GDP <- GDP%>%
filter(country_name %in%
(GDP %>% distinct(country_name, mean_rank) %>%
top_n(10, -mean_rank) %>%
pull(country_name)))
GDP <- GDP %>%
# rereank
group_by(year) %>%
mutate(rank = rank(-GDP_per_capita, ties.method = "random")) %>%
ungroup() %>%
select(-mean_rank)# code to get codes of the countries (to utilize geom_flag())
country_2_letters <- countrycode(
GDP$country_name %>%
unique() %>%
sort(),
origin = "country.name", destination = "genc2c") %>%
tolower() %>%
set_names(GDP$country_name %>% unique() %>% sort())
GDP <- GDP %>%
mutate(country_code = country_2_letters[country_name])GDP_plot <-
ggplot(data=GDP, aes(year, rank, group=country_name, color= country_name,
fill= country_name)) +
geom_bump(smooth=10, size=1.5, lineend= "round", alpha=.8) +
geom_flag(data = GDP %>% filter(year==min(year)), aes(country=country_code),
size=8, color="black") +
geom_flag(data = GDP %>% filter(year==max(year)), aes(country=country_code),
size=8) +
scale_color_paletteer_d("tidyquant::tq_light") +
# label for each contry
geom_text(data = GDP %>% filter(year == max(year)),
aes(label = country_name),
color = "gray16",
nudge_x = .51,
hjust = 0,
size = 3,
fontface = 2) +
xlim(c(2000, 2022)) +
scale_y_reverse(breaks= 1:10) +
theme_classic() +
theme(legend.position = "none",
) +
labs(x = NULL,
y = NULL,
title = "GDP per capita ranking",
subtitle = "For 10 countries that were ranked the highest in 2000-2020")GDP_plot
Luxembourg comes top each year, we can also see the gradual rise of Ireland.
The correlation between active youth and GDP
The evaluation is for Europe only.
indicators_eu <- indicators_base %>%
mutate(
continent_name = countrycode(
sourcevar = indicators_base$country_name,
origin = "country.name",
destination = "continent")) %>%
dplyr::relocate(continent_name, .after=country_name) %>%
dplyr::filter(continent_name == "Europe")reason_s= "Share of youth not in education, employment or training, total (% of youth population)"
youth_GDP_eu <- indicators_eu %>%
filter(series %in% c(reason_s, "GDP per capita (current US$)")) %>%
select(country_name, series, "1970": "2020")%>%
pivot_longer(cols = "1970":"2020", names_to = "year", values_to = "value") %>%
pivot_wider(names_from = series, values_from = value) %>%
rename(reason_val= reason_s , GDP_p_c = "GDP per capita (current US$)") %>%
mutate(year = as.integer(year),
reason_val = as.double(reason_val),
GDP_p_c = as.numeric(GDP_p_c))youth_GDP_eu_2017 <- youth_GDP_eu %>%
filter(year == 2017) %>%
filter(country_name != "Russian Federation") %>%
drop_na(reason_val, GDP_p_c)youth_GDP_plot <- ggplot(data= youth_GDP_eu_2017, aes(x= reason_val, y = GDP_p_c, label=country_name)) +
#geom_point( size= 0, alpha= .6) +
geom_label_repel(seed = 23, size = 3, max.overlaps = 8, segment.color= "grey30",
fontface= "bold",
fill = ifelse(youth_GDP_eu_2017$country_name == "Italy","#E31A1C" ,"#18BC9C"), color = "white") + # color from cb set 3
theme_minimal() +
labs(title = "Influence of youth not taking education on GDP per capita",
subtitle = "Chosen European countries",
caption = "Data from 2017") +
xlab("% of lazy youth") +
ylab("GDP per capita")
youth_GDP_plot
The chart illustrates a strong correlation between the share of youth not in education, employment or training and the GDP per capita. It also shows one deviation - Italy (marked red) has relatively high GDP/capita considering how many young people from this country do not take up any education or work.
Rural population change
The following map illustrates the change of people living in cities in years 1970-2020.
On the next visualization we will see how the urbanization is correlated with GDP per capita.
world <- map_data("world")
feature_map <- "Urban population (% of total population)"
data_map <- indicators_base %>%
filter(series == feature_map ) %>%
filter(country_name != "World") %>%
select(country_name, country_code,
'1970', '1980', '1990', '2000', '2010', '2020') %>%
# single row - single observation - using tidyr
pivot_longer(cols = 3:8, names_to= 'year', values_to = 'feature_val') %>%
# transform datatypes
mutate(feature_val = as.numeric(feature_val),
year = as.numeric(year)) %>%
mutate(country_name = recode(str_trim(country_name),
"United States" = "USA",
"United Kingdom" = "UK",
"Russian Federation" = "Russia",
"Korea, Dem. People's Rep." = "South Korea",
"Congo Dem. Rep." = "Democratic Republic of the Congo" ,
"Congo Rep." = "Republic of Congo",
"Egypt, Arab Rep." = "Egypt"))# merging the datasets (polygons and values)
data_map <- inner_join(world, data_map, by= c("region" = "country_name"), keep= TRUE)## Removing unnecessary elements
plain <- theme(
axis.text = element_blank(),
axis.line = element_blank(),
axis.ticks = element_blank(),
panel.border = element_blank(),
panel.grid = element_blank(),
axis.title = element_blank(),
panel.background = element_rect(fill = "white"),
plot.title = element_text(hjust = 0.5),
)
world_map_static <-
ggplot(data = data_map, aes(x=long, y=lat, group=group)) +
coord_fixed(1.3) +
geom_polygon(aes(fill = feature_val)) +
scale_fill_distiller(type= "seq", palette = 14, direction = 1, na.value="grey33", values = c(0,1), ) +
plain +
labs(fill="Urbanization (%)")
# world_map_static# animation
anim_pop <- world_map_static + transition_time(year, ) +
ggtitle("Urban population (% of total population)",
subtitle ='Year: {frame_time}')
num_years <- 6
gif_anim <- animate(anim_pop, nframes = num_years, fps = 0.8)gif_anim
Urbanizationâs influence on GDP
On the chart we can see the correlation between the urbanization level and the GDP per capita. We also show the percentage of population that reaches the age of 65 as a feature that is strongly related with GDP per capita.
GDP per capita for each country is represented by the size of a bubble.
feature_x <- "Employers, total (% of total employment) (modeled ILO estimate)"
feature_y <- "Survival to age 65, male (% of cohort)"
feature_z <- "GDP per capita (current US$)"
feature_x <- "Urban population (% of total population)"
data_buble <- indicators_base %>%
mutate(continent_name = countrycode(
sourcevar = indicators_base$country_name,
origin = "country.name",
destination = "continent")) %>%
filter(series %in% c(feature_x, feature_y, feature_z)) %>%
select(country_name, continent_name, series, "1970":"2020") %>%
pivot_longer(cols = "1970":"2020", names_to = "year", values_to= "value") %>%
pivot_wider(names_from = series, values_from = value) %>%
rename(feature_x_val = feature_x, feature_y_val = feature_y, feature_z_val = feature_z) %>%
mutate(year = as.integer(year),
feature_x_val = as.numeric(feature_x_val),
feature_y_val = as.numeric(feature_y_val),
feature_z_val = as.numeric(feature_z_val))data_bubble_2017 <- data_buble %>%
filter(year == 2017) %>%
drop_na()bubble_chart <- ggplot(data = data_bubble_2017, mapping = aes(x = feature_x_val, y = feature_y_val)) +
geom_point(aes(size = feature_z_val, color = continent_name, text = paste0(country_name,"\n", "GDP p.c.:", round(feature_z_val))), alpha = .67, ) +
#scale_color_brewer(palette = "") +
#scale_colour_paletteer_d("ggsci::category10_d3" ) +
scale_color_paletteer_d("tidyquant::tq_light", ) +
theme_minimal() +
labs(
caption = "Data from 2017",
color= "Continent") +
xlab("Urban population (% of total population)") +
ylab("Survival to age 65, male (% of cohort)") +
guides(size = FALSE)
ggplotly(bubble_chart, tooltip = c("text", "group"))We can clearly see that for African countries the urbanization ratio is lower than for the rest of the world. Both indicators seem to be strongly correlated with the GDP per capita.
In top-right corner
Greenhouse gas emissions analysis
Top emitters
This animation presents the largest greenhouse gas emitters in years 1970-2020.
wdi_cleaned <- wdi %>%
rename_at(vars(oldnames), ~ newnames)
greenhouse <- wdi_cleaned %>%
filter(`Series Name` == "Total greenhouse gas emissions (kt of CO2 equivalent)") %>%
filter(!`Country Name` %in% c("World","Upper middle income","High income","Low & middle income","Low income","Middle income","Lower middle income")) %>%
mutate(`Country Code` = countrycode::countrycode(`Country Code`, "iso3c","iso2c")) %>% #%>%
select(`Country Name`,`Country Code`,`1970`:`2020`) %>%
gather(year,value,`1970`: `2020`) %>%
mutate(value=replace(value,value=='..',0)) %>%
transform(value=as.numeric(value)) %>%
mutate(`Country Name`= `Country.Name`) %>% select(!`Country.Name`) %>%
group_by(year) %>%
mutate(rank = rank(-value),
Value_lbl = paste0(" ",round(value))) %>%
filter(rank <=10) %>% mutate(year = as.numeric(year))
greenhouse$Country.Code = tolower(greenhouse$Country.Code)staticplot = ggplot(data=greenhouse,aes(rank,value, group=`Country Name`, color= `Country Name`, fill= `Country Name`)) +
geom_col(aes(x = rank, y = value), fill = "azure3", color = "black") + # Columns
theme_classic() +
theme(legend.position = "none",axis.text.y = element_blank()
,plot.margin = margin(0,2,0,3,"cm")) +
geom_text(aes(label = `Country Name`,x = rank,
y = -200),
color = "gray16",hjust=1) +
coord_flip(clip = "off", expand = FALSE) + # Flip
geom_flag(aes(x = rank, y =value, country = `Country.Code`), size = 10) + # Flags
scale_y_continuous(labels = scales::comma) + # Format y-axis values
scale_x_reverse() + # Highest values on top
transition_states(year, transition_length = 4, state_length = 1) + # Animate
scale_color_paletteer_d("tidyquant::tq_light") +
labs(x = "", y = "kt of CO2",title = "Biggest greenhouse gas emitters (kt of CO2 equivalent)",subtitle= "years 1970-2020") # object `year` not foundstaticplot
Animation shows that before 2000`s the biggest emitters were USA and Russia, with China taking a huge lead after. Over time European countries slowly decline their emissions whilst developing and manufacturing focused countries increased them.
Carbon emissions by fuel state of matter in EU
We differentiate three types:
solid : anthracite, coal, long lasting ovals, coke
liquid : petroleum based fuels, natural gas liquids, biofuels
gas : LNG, natural gas, producer gas, coal gas
The animation presents changes in EU member statesâ fuel type use in years 1992-2015
#prepare emissions by type
#note years (1992-2015) chosen for data integrity
gas_emissions <- wdi_cleaned %>%
filter(`Series Name` == "CO2 emissions from gaseous fuel consumption (kt)") %>%
gather(year,gas,`1992`:`2015`) %>%
select(`Country Name`,`Country Code`, `year`,`gas`) %>%
mutate(gas=as.numeric(replace(gas,gas=='..',0)))
liquid_emissions <- wdi_cleaned %>%
filter(`Series Name` == "CO2 emissions from liquid fuel consumption (kt)") %>%
gather(year,liquid,`1992`:`2015`) %>%
select(`Country Name`,`Country Code`, `year`,`liquid`) %>%
mutate(liquid=as.numeric(replace(liquid,liquid=='..',0)))
solid_emissions <- wdi_cleaned %>%
filter(`Series Name` == "CO2 emissions from solid fuel consumption (kt)") %>%
gather(year,solid,`1992`:`2015`) %>%
select(`Country Name`,`Country Code`, `year`,`solid`) %>%
mutate(solid=as.numeric(replace(solid,solid=='..',0)))
#join into one data frame
all_emissions <- left_join(gas_emissions,liquid_emissions,by=c('Country Name','Country Code','year')) %>%
left_join(.,solid_emissions,by=c('Country Name', 'Country Code','year')) %>%
mutate(sum=100-rowSums(across(gas:solid))) #load EU countries` borders
get_eurostat_geospatial(resolution = 10,
nuts_level = 0,
year = 2016)## Simple feature collection with 37 features and 11 fields
## Geometry type: MULTIPOLYGON
## Dimension: XY
## Bounding box: xmin: -63.08825 ymin: -21.38917 xmax: 55.83616 ymax: 71.15304
## Geodetic CRS: WGS 84
## First 10 features:
## id LEVL_CODE NUTS_ID CNTR_CODE NAME_LATN
## 1 AL 0 AL AL SHQIPĂRIA
## 2 AT 0 AT AT ĂSTERREICH
## 3 BE 0 BE BE BELGIQUE-BELGIĂ
## 4 NL 0 NL NL NEDERLAND
## 5 PL 0 PL PL POLSKA
## 6 PT 0 PT PT PORTUGAL
## 7 DK 0 DK DK DANMARK
## 8 DE 0 DE DE DEUTSCHLAND
## 9 EL 0 EL EL ELLADA
## 10 ES 0 ES ES ESPANA
## NUTS_NAME MOUNT_TYPE URBN_TYPE
## 1 SHQIPĂRIA 0 0
## 2 ĂSTERREICH 0 0
## 3 BELGIQUE-BELGIĂ 0 0
## 4 NEDERLAND 0 0
## 5 POLSKA 0 0
## 6 PORTUGAL 0 0
## 7 DANMARK 0 0
## 8 DEUTSCHLAND 0 0
## 9 <U+0395><U+039B><U+039B><U+0391><U+0394><U+0391> 0 0
## 10 ESPANA 0 0
## COAST_TYPE FID geometry geo
## 1 0 AL MULTIPOLYGON (((19.82698 42... AL
## 2 0 AT MULTIPOLYGON (((15.54245 48... AT
## 3 0 BE MULTIPOLYGON (((5.10218 51.... BE
## 4 0 NL MULTIPOLYGON (((6.87491 53.... NL
## 5 0 PL MULTIPOLYGON (((18.95003 54... PL
## 6 0 PT MULTIPOLYGON (((-8.16508 41... PT
## 7 0 DK MULTIPOLYGON (((14.8254 55.... DK
## 8 0 DE MULTIPOLYGON (((8.63593 54.... DE
## 9 0 EL MULTIPOLYGON (((29.60853 36... EL
## 10 0 ES MULTIPOLYGON (((4.28746 39.... ESSHP_0 <- get_eurostat_geospatial(resolution = 10,
nuts_level = 0,
year = 2021)
EU27 <- eu_countries %>%
filter(code != 'UK') %>%
select(geo = code, name)
SHP_27 <- SHP_0 %>%
select(geo = NUTS_ID, geometry) %>%
inner_join(EU27, by = "geo") %>%
arrange(geo) %>%
st_as_sf()
# move into one iso2c format -> used in eurostat
all_emissions_cleaned <- all_emissions %>% mutate(`Country Code` = countrycode::countrycode(`Country Code`, "iso3c","iso2c"))
europe_emmissions <- all_emissions_cleaned %>%filter(`Country Code` %in% SHP_27$geo)
# join geospatial data with emissions
europe_emmissions_geo <- left_join(europe_emmissions,SHP_27,by=c("Country Code" = "geo"))
# calculate geographical midpoints of countries
centroids <- europe_emmissions_geo$geometry %>% st_centroid() %>%
# this is the crs from d, which has no EPSG code:
st_transform(., '+proj=longlat +ellps=GRS80 +no_defs') %>%
# since you want the centroids in a second geometry col:
st_geometry() %>% st_as_sf() %>% transform(country = europe_emmissions_geo$`Country Name`) %>%
mutate(x=replace(x,country=="France",st_point(c(2,46)))) %>% #France borders include overseas territories
mutate(x=replace(x,country=="Czech Republic",st_point(c(15.1,49.9)))) %>% # correct error in calculation
mutate(x=replace(x,country=="Slovak Republic",st_point(c(19.78,48.7)))) # correct error in calculation
#adjust data type to fit the scatterpie requirements
lats = c()
longs = c()
for (center in centroids$x) {
lats <- rbind(lats,st_bbox(center)$xmin)
longs <- rbind(longs,st_bbox(center)$ymin)
}
lats[is.na(lats)] <- 0
longs[is.na(longs)] <- 0
europe_emissions_lat_long <- europe_emmissions_geo %>% select(`Country Name`,year,gas,liquid,solid,geometry) %>% transform(lats =as.numeric(lats)) %>% transform(longs=as.numeric(longs))
# necessary to generate and save plots for each year
# using gganimate throws errors when trying to animate both polygons and pie charts
for (yr in europe_emissions_lat_long$year %>% unique()) {
emissions_for_year <- europe_emissions_lat_long %>% filter(year==yr)
plot <- ggplot(emissions_for_year)+ggtitle(paste("Carbon emissions by type of fuel\n year:",yr)) + geom_sf(aes(geometry=geometry)) +
scale_x_continuous(limits =c(-10, 35)) +
scale_y_continuous(limits = c(33, 68)) + theme_void() +geom_scatterpie(aes(x=lats,y=longs,r=0.85),data=emissions_for_year,cols=c("gas","solid","liquid"))
print(plot)
ggsave(paste(trimws(yr),"plot.png",sep=''),bg="#FFFFFF")
dev.off()
}
# create a gif from generated images
library(gifski)
png_files <- list.files(".", pattern = ".*png$", full.names = TRUE)
#png_files
animation <- gifski(png_files, gif_file = "animation.gif", width = 1600, height = 1200, delay = 0.5)
#animation
For 20 years, most EU countries did not change their sources of fuel by a large margin. The most noticeable are Baltic States which increased their reliance on gas from liquid fuels, Poland`s solid fuel use slowly declining and Luxembourg cutting down solid fuels completely.
Manufacturing vs emissions vs GDP per capita
The following bubble chart shows correlations between three indicators % of manufacturing in economy, greenhouse gas emissions and GDP per capita
feature_x <- "GDP per capita (current US$)"
feature_z <- "Total greenhouse gas emissions (kt of CO2 equivalent)"
feature_y <- "Manufacturing, value added (% of GDP)"
emissions <- wdi_cleaned %>%
mutate(continent_name = countrycode(
sourcevar = wdi_cleaned$`Country Name`,
origin = "country.name",
destination = "continent")) %>%
filter(`Series Name` %in% c(feature_x, feature_y, feature_z)) %>%
select(`Country Name`, continent_name, `Series Name`, "1970":"2020") %>%
pivot_longer(cols = "1970":"2020", names_to = "year", values_to= "value") %>%
pivot_wider(names_from = `Series Name`, values_from = value) %>%
rename(feature_x_val = feature_x, feature_y_val = feature_y, feature_z_val = feature_z) %>%
mutate(year = as.integer(year),
feature_x_val = as.numeric(feature_x_val),
feature_y_val = as.numeric(feature_y_val),
feature_z_val = as.numeric(feature_z_val))
emissions_2015 <- emissions %>%
filter(year == 2015) %>%
drop_na()
bubble_chart <- ggplot(data = emissions_2015, mapping = aes(x = feature_x_val, y = feature_y_val)) +
geom_point(aes(size = feature_z_val, color = continent_name, text = paste0(`Country Name`,"\n", "kt of CO2: ", round(feature_z_val))), alpha = .67, ) +
scale_color_paletteer_d("tidyquant::tq_light", ) +
theme_minimal() +
labs(
caption = "Data from 2015",
color= "Continent") +
xlab("GDP per capita (current US$)") +
ylab("Manufacturing, value added (% of GDP)") +
guides(size = FALSE)
ggplotly(bubble_chart, tooltip = c("text", "group"))In general we can observe that higher % of manufacturing in economy, loosely correlate with higher greenhouse gas emissions, presumably because of increased need for energy. One outlier - Ireland seems to defy that rule, however that is because of Apple moving headquarters to Ireland . We can also see that the richest countries with exception to USA do not produce a lot of emissions. One of the reasons why may be a type of products manufactured, three largest companies in Europe belong to automotive industry